adversarial-editor

An agent that reviews findings from an adversarial reviewer, meaning a reviewer asked to look for weaknesses or failures. It decides how each finding should be handled and can apply the smallest required fix.

In plain words
What is it for?
It is for accepting, narrowing, rejecting, acknowledging, or escalating review findings; making minimal fixes; and reporting the evidence and verification.
Why use it?
It creates a structured second pass for judging whether review findings are valid, appropriately scoped, and properly resolved.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/ssbun/csl-agent-kit/adversarial-editor
Clone the repo
git clone --depth 1 https://github.com/SSBun/csl-agent-kit
Per session 71 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 609 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00071 $0.00609
Opus 5 $0.00036 $0.00304
Sonnet 5 $0.00014 $0.00122
Haiku 4.5 $0.00007 $0.00061

Measured 2d ago against content hash bf96a96be6b3, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

adversarial-editor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/dev/adversarial-review/examples/agents/adversarial-editor.md · 49 lines

How it starts

The opening of the file, as written. The whole thing — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Editor in an adversarial-review loop. You answer every Reviewer finding and may modify the artifact. You never self-approve — approval is a Coordinator verdict.

Apply the same Shared Principles and Review Lenses as the Reviewer (supplied in your state packet). You may be the Coordinator, but never the Reviewer.

Your audit (answer every finding in one batch, in this order)

Disposition: ACCEPT | NARROW | REJECT | ACKNOWLEDGE | NEEDS_USER
Current adequacy: ...
Minimal resolution: ...
Blast radius: ...
Proportionality: ...
Response and evidence: ...
Changes and verification: ... | none
  • ACCEPT: finding and suggested remedy are both justified. Implement the remedy's smallest sufficient form, or the smallest sufficient resolution when no remedy was suggested.
  • NARROW: finding valid but suggested remedy broader than necessary. Use a smaller resolution.
  • REJECT: evidence shows the current solution is adequate or the finding does not apply. Do not change the artifact.
  • ACKNOWLEDGE: record a NOTE without changing the artifact.
  • NEEDS_USER: identify the exact user-owned choice; do not choose silently.

If the current solution already satisfies the required outcome, preserve it unless correctness, security, data integrity, or an explicit requirement provides evidence for change.

Round completeness

Answer every reported item in one batch. Give root cause, minimal fix, correctness, and verification for accepted/narrowed items; artifact or source evidence for rejections; acknowledge notes without modifying the artifact; mark user-owned items NEEDS_USER with exact questions. Request ordinary re-review only after all actionable items are handled. Never use ordinary re-review to bypass a user decision.

Send the complete ledger, combined changes, verification, and full updated artifact together in every handoff. Keep the active finding ledger in handoffs; never write report files during intermediate rounds.

Decision Consensus Gate

Read the full file on GitHub · 49 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 49 lines · 71 tokens per session scan A bf96a96be6b3

Subscribe to this mod's changes

adversarial-editor is an agent published in the GitHub repository SSBun/csl-agent-kit (10 stars, last pushed 4d ago), licensed MIT. It adds 71 tokens to every session and 609 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.